Is there anyone who can help me understand the Sox/R8Brain upsampling options?
I battled my way through the āinfoā button on the settings page and still couldnāt see when one would choose Sox and when R8Brain. And which options for each.
I have similarly waded through the HQ Player documentation without understanding it.
I watched Passion for Soundās video on HQ Player settings, and I can see he doesnāt understand it either.
So I have done due diligence on trying to understand all of this.
You can think of this in three partsā¦ Up/Down sample, phase and shape/rolloffā¦
Archimago does a good job of explaining and even has some pictures that may helpā¦
Everyoneās kit, room, ears and brain are differentā¦ No single choice of parameters will be loved by everyoneā¦ Trust your own ears and donāt worry about what others think when it comes to your own personal taste in soundā¦
All solid points given here ( including @Jacob ) Itās going to come down to your personal preferences, ears and gears. Iād suggest taking a screenshot of the default settings and try both options out with and without upsampling. Your equipment may or may not be capable of high upsampling to DSD perhaps. Have some good fun with it
Best luck with it.
Damien has told me that they are getting best results with r8Brain primarily due to the more current codeā¦ So I use r8Brain.
r8Brain is simple to implement if you understand three important things: First is to understand what Nyquist Frequency is, and the ramifications of where the Nyquist filter is appliedā¦ And, what does āStopbandā mean in a filter design and the ramifications of the stopband filter attenuation settingā¦ And finally the difference between minimum phase and linear phase filter affects on the signalā¦
Monkey around with the settings to ge a feel for how it affects the sound-quality of your files.
Like Ddude003 recommends:
However, It is always good to get a handle on what is going on in those algorithms/filtersā¦
I would advise not to pay much attention to this, itās not something to focus on, use tidal or qobuz, and improve your equipment and you will be fine.
Upsampling will be done to PCM format files regardless. If you send bit perfect PCM to virtually any consumer DAC, with only a handful of exceptions, it will be upsampled inside the DAC and converted to DSD or a format similar to DSD. So your choice isnāt so much whether or not to upsample, but where. Your computer will pretty much always have more resources for this conversion, and be able to do it in a more sophisticated manner.
R8Brain is considered by the developer, Damien, to be the better upsampling algorithm. Subjectively I feel the same way, but listen to both and make your own choice.
The best conversion settings depend on your DAC and speakers. What are they?
Yesā¦
In the case of the iFi DAC that you have, with the Burr-Brown chipset , all 1-bit PDM (DSD) signal is routed directly to the output circuitryā¦, DSDxxx never gets routed through the PCM to multi-bit SDM (Sigma Delta Modulation) up-sampling circuity of the chipset before routing to the output circuitry as 1-bit PDM signalā¦ All PCMxxx will be routed through the SDM process, but the higher-the sample-rate, the less conversion processing the chipset is taxed withā¦ The interpretation of the resulting sound, is still subjective, however.
You would want to upsample to the max your DAC can input. Strangely Iāve seen DSD512 and DSD256 for this unit, but whichever it is, thatās what you want.
You probably want to use r8brain, which is generally thought to be the better upsampling/filtering software.
Among audiophile speakers, as far as I know only Vandersteens use linear phase crossovers. Whatās that mean? Phase is timing. Linear phase means all frequencies take the same amount of time to pass through the crossover. (Why doesnāt everyone have linear phase crossovers? Because it causes unavoidable bumps, small but there, in frequency response at the crossover points.) You donāt have Vandersteens. So unless you have room equalization/correction software, you can experiment with settings other than linear phase and see what you like. If you do use such software, you want to keep the timing as is, so use linear phase. (At other phase settings, some frequencies pass through the filter slower than others, which is called āgroup delay.ā)
You probably want to start off with the cutoff set as high as possible. No sense filtering out the highest harmonics.
You likely want to use the maximum setting for anti-aliasing, or whatever the setting that governs how sharply the filter cuts is called. Thatās noise and distortion up there, no sense letting it through.
For the modulator setting you want B7, B8, or C (Iāve been told C and B7 are identical). These let through the least noise and distortion.
(Note when I talk about letting through noise and distortion, the levels from most of the modulators may be inaudible. But why not set things up to be as accurate as possible?)
Thanks, Iāve never played with the settings, they are default in R8Brainā¦ But Iām willing to tryā¦
The DAC does to 256, but being on a Mac, the 256 is greyed out, I suppose as per the p.o.p rules.
these are the settings.
The one setting you might want to change is volume reduction before DSD upsampling, setting it to reduce volume by something like 3dB.
Some engineers like to record āhot.ā When you go above max volume (0dB attenuation) while recording digitally, you get distortion. The recording will reproduce this distortion through āintersample oversā (the waveform goes above the 0dB max between samples). The safe volume reduction setting avoids these intersample overs.
Hi @Reynaldo , itās been forever since I first used upsampling, so no idea of the defaults. But my settings are upsampling with r8brain to DSD512, bandwidth 99.5, stop band attenuation 218, linear phase, B 8th order modulator (I sometimes switch between this and C, no particular reason, just playing around and if one subjectively strikes me as sounding better that dayā¦), volume reduction -3dB.
Since Iām upsampling to the max input rate of the NEO iDSD, I donāt think the internal DAC filter selection matters at all - the DSD512 input should bypass the DAC chip and its filters and modulators entirely. Out of an excess of caution I have the filter set to Bit Perfect so the DAC is told not to try to employ its digital filtering, but I doubt this is actually needed.
I utilize the āAā type 8th order SDM filterā¦ (I believe this is an apodizing filter) in concert with a FIR filter in the DAC (Fc = 90kHz, Gain = +0.3dB)
Youāll see that I donāt reduce the volumeā¦ ISPās (Inter-Sample Overs) are real, however, I rarely detect these spurious events, When I down-convert DSDxxx to DXD 352.8KHz for HRTF processing, I apply ISP elimination processingā¦ Generally, if you are using plug-in processing before modulating to DSDxxx there is good reason to reduce the volume before DSD upsampling by at least 3dBā¦ Over the years of having this function available to apply to the file, lately, I have found no reason to reduce the volume, in conjunction with the Abbey Road Studio 3 HRTF virtualization AU plug-in.
The DoP setting is found in the DAC Input section of the settingsā¦
The Stop Band is where the Low-pass filter cutoff is applied at the top-end and may also be applied to the low-end High Pass cut-off frequencyā¦ I donāt know if the low-end cut-off function is applied in the r8Brain algorithmā¦ This is most likely associated to the Filter Type that is being appliedā¦ Iāve found that small adjustments in the Stop Band Fc makes a big difference in the high-end energyā¦ Tune to taste.
Come to think about this a bit moreā¦
I donāt remember if r8Brain actually allows the choice of Filter Typesā¦ What is showing in the screen-shot is what I was using in SoXā¦ For some reason I sort of remember Damien saying something about thisā¦ This needs clarificationā¦
But my settings are upsampling with r8brain to DSD512, bandwidth 99.5, stop band attenuation 218, linear phase, B 8th order modulator (I sometimes switch between this and C, no particular reason, just playing around and if one subjectively strikes me as sounding better that dayā¦), volume reduction -3dB.